> This is the caching API we've always wanted Workers to have. Here's why it took us this long
I was looking forward to the "why it took us this long" explanation but it wasn't explicitly spelled out. Any Cloudflare staff here able to expand on that?
(The article does a good job of showing how many different smart design decisions went into this, but given caching is core to what a Cloudflare does I'm still a little surprised it took 9 years to get here!)
Huge props to just sticking with the HTTP spec on this one with `Cache-Control` headers with `stale-while-revalidate` support. It's amazing how many other providers mess that up.
On top of that the cache tags are a slick way to do invalidation. This looks like a great product.
People that don't like writing now get to write by offloading it to an LLM, and this is the result. I miss the world where articles were mostly written by people who had the interest and patience to do it.
I was a bit confused what this adds other than just standard CDN-Cache-Control page caching that we do now. Some quirks I've found;
- You still get billed per request, whether the request hits cache or not (but don't get billed for CPU time)
- You now get billed for static asset requests! This makes no sense to me. "One thing to watch: when caching is enabled, requests that are normally free — static asset requests and worker-to-worker invocations through service bindings or ctx.exports — are billed at the standard request rate, because each one now consults the cache in front of your Worker." Yeah that sounds like a bug that just happens to generate them more money.
- The cache key automatically has the worker deployment version, so even gradual deployments populate their own cache which is nice
- It seems like you can set a totally custom cache key? But that was previously Enterprise only, can't see if that's still the case here.
The previous Worker in-front of the cache never made sense to my old school proxy in-front of the appserver mindset. Already using this to speed up a tool. Nice work.
Amazing, exactly what Workers lacked, I was quite annoyed that a worker would spin up when 99% of my requests then return a cache. Waste of 3ms, times millions.
Reinventing the web, one service at a time. Is it really that hard to build things like this on a per-case base, small scale, all on a bare bones VPS? Do we need a global gateway to help us not screw up the basics? Service looks nice but makes me sad somehow.
As a DevOps engineer, if you learn to implement CF Workers with Claude Code's Cloudflare skill, you will no longer be burdened by this question. I'm running a Docker swarm on my homelab, a Docker swarm on a EC2 instance, but there's no longer a single web service in either of them. Workers is insanely good.
When Workers Cache is enabled, every cacheable request to your Worker hits Cloudflare's cache first. If there's a fresh cached response, Cloudflare returns it directly — your Worker doesn't run, and you don't pay CPU time for it. On a miss, your Worker runs, and if your response is cacheable, Cloudflare stores it for the next request. The next request from anywhere on Earth can be served straight from cache.
Incredible! This is why I shoehorn all my server side usecases on to the Workers Platform. Cloudflare, since 2020 when I first went all-in, has consistently shipped features that reduce bills significantly (except for 2023 Workers usage model changes). In one case, when they shipped free Snippets (Workers but 32kb code size & 5s CPU time) for Pro accounts ($200/yr), our bills went from £15k+ to £0.
I know about the infamous "Enterprise plan" (especially, when your bandwidth is as high as ours in 100s of TBs) and know of at least one other tech shop that was required to pay for it ... but we haven't got that sales call, yet.
Our bandwidth is very high, we constantly get invited onto the call with their team, but after talking with them a few times it makes absolutely zero sense for us to have a committed spend, all the stuff I needed an account manager for in GCP/AWS just doesn't exist in CF. Support wise I imagine if it's broken for us it's also broken for 2 million other people so... yeah... Thanks CF!
One of the recent AI tells other than em dash is the excessive usage of hyphenated words:
> multi-tenant-safe cache keys
> on a server-rendered app
> byte-for-byte identical (classic)
> gets a cache-speed response
> cached-file-extensions list
Honestly, this is terrible. I had to add a "use simple words only, don't hyphenate unnecessarily" warning to my Claude config. After a full day of work, having to read these Claudisms all the time make a noticeable difference on how tired you get. It gets even worse when Claude starts to make up its own vocabulary.
Fuck, I spent all these years developing a thoughtful writing style that leaned toward clarity for the reader, even if it meant extra work to achieve precision, or adding affordances like “excessive” hyphenation, and now I guess have to learn to write worse.
I think it’s exceedingly unlikely for a good-faith reader to mistake good-faith human writing for AI writing.
Even if you use em dashes and a few phrases that have become associated with AI writing, there’s still an unmistakeable sense of how much effort was put into the writing.
But I suppose there might be naive readers who don’t know how to spot this effort and would false positive on em dashes or supposed AI phrases.
There is a world of difference between well-written human text and sloppy walls of AI-generated text. There's nothing wrong in using hyphenations or emdashes -- I use them myself! That's not the point of my comment.
Whether we like it or not, em dashes are effectively verboten in online discussions and blog posts if you want people to take you seriously. If the idea that excessive hyphenation is an AI tell gains traction, it too will become impossible to use without ruining your credibility.
What does it mean when prompting SoTA LLMs prone to slop to be concise and precise, with respect to context at hand, not work at all? Anyone benchmarking that?
Don't change. The homogenous way LLMs write is just tiresome and boring, like if every movie stared Ryan Reynolds - an actor famous for having no range. Ryan Reynolds is enjoyable to watch on occasion, but I don't want everything I watch to be Ryan Reynolds.
They could just feed an llm a small corpus of past human authored posts from their site, and have the LLM rewrite it in a style matching style, and it would likely turn out pretty great.
"slop" doesn't mean "AI generated content", it means bad content, a waste of the reader's time. Grantparent's implication is that your comment was bad content, not that it was AI generated bad content.
Not the original commenter; but, at least for me, the idea is that when it’s written by humans we know that effort and care were put into communicating the news. Otherwise they could post a link to the docs and we could ask my flavor of LLM to summerize. No need for extra filler content. That why it’s slop and it’s different.
I don’t have access to my flavor of LLM on the train nor the time or budget to have it do the research and summary for me
Why are we all of a sudden pretending like pre-LLM era blogs were these pristinely well written pieces of art or even that effort and care was put into them? In most cases they were significantly less coherent and incomplete. Don’t get me started on the mess that was the communication of this particular company or one of their competitors like AWS.
Because we add humans enjoy variety. I read for entertainment, even technical posts like this one that I have no use for. I often trying to think about what the author may have been thinking when writing, why they introduced concepts in a specific order, what ideas might they have omitted, etc. It's personal and enjoyable. But now, when I detect the familiar writing style of what seems to be a gpt 5 model, that "parasocial" connection dies.
The LLM explained the core concept and features very well. But it was dull and boring to me, as I already have to read this writing style at work pretty much all day every day.
>I don’t have access to my flavor of LLM on the train nor the time or budget to have it do the research and summary for me
Aren't you gonna let the LLM develop for you anyway? Why bother writing and reading a post at all?
> pretending like pre-LLM era blogs were these pristinely well written pieces of art
The point is the effort and care that the writer puts which differentiates it from automatically generated text. That matters because a human can sympathize and that leads to better understanding and greater connection. That's why a post is written.
> Don’t get me started on the mess that was the communication of this particular company or one of their competitors like AWS.
And we criticize those as well. Nothing's changed. Yesterday's bad content is today's slop (plus a mind boggling amount of investment, corruption and environmental side effects).
I was looking forward to the "why it took us this long" explanation but it wasn't explicitly spelled out. Any Cloudflare staff here able to expand on that?
(The article does a good job of showing how many different smart design decisions went into this, but given caching is core to what a Cloudflare does I'm still a little surprised it took 9 years to get here!)
On top of that the cache tags are a slick way to do invalidation. This looks like a great product.
- You still get billed per request, whether the request hits cache or not (but don't get billed for CPU time)
- You now get billed for static asset requests! This makes no sense to me. "One thing to watch: when caching is enabled, requests that are normally free — static asset requests and worker-to-worker invocations through service bindings or ctx.exports — are billed at the standard request rate, because each one now consults the cache in front of your Worker." Yeah that sounds like a bug that just happens to generate them more money.
- The cache key automatically has the worker deployment version, so even gradual deployments populate their own cache which is nice
- It seems like you can set a totally custom cache key? But that was previously Enterprise only, can't see if that's still the case here.
A big worry was always "why does workers sit in front of my cache? that's a waste of an invocation if i'm returning a cached result"
Thanks CF.
I know about the infamous "Enterprise plan" (especially, when your bandwidth is as high as ours in 100s of TBs) and know of at least one other tech shop that was required to pay for it ... but we haven't got that sales call, yet.
> multi-tenant-safe cache keys
> on a server-rendered app
> byte-for-byte identical (classic)
> gets a cache-speed response
> cached-file-extensions list
Honestly, this is terrible. I had to add a "use simple words only, don't hyphenate unnecessarily" warning to my Claude config. After a full day of work, having to read these Claudisms all the time make a noticeable difference on how tired you get. It gets even worse when Claude starts to make up its own vocabulary.
Even if you use em dashes and a few phrases that have become associated with AI writing, there’s still an unmistakeable sense of how much effort was put into the writing.
But I suppose there might be naive readers who don’t know how to spot this effort and would false positive on em dashes or supposed AI phrases.
Responding to alleged slop with more slop doesn’t decrease the total amount of slop on the internet.
Why are we all of a sudden pretending like pre-LLM era blogs were these pristinely well written pieces of art or even that effort and care was put into them? In most cases they were significantly less coherent and incomplete. Don’t get me started on the mess that was the communication of this particular company or one of their competitors like AWS.
The LLM explained the core concept and features very well. But it was dull and boring to me, as I already have to read this writing style at work pretty much all day every day.
Aren't you gonna let the LLM develop for you anyway? Why bother writing and reading a post at all?
> pretending like pre-LLM era blogs were these pristinely well written pieces of art
The point is the effort and care that the writer puts which differentiates it from automatically generated text. That matters because a human can sympathize and that leads to better understanding and greater connection. That's why a post is written.
> Don’t get me started on the mess that was the communication of this particular company or one of their competitors like AWS.
And we criticize those as well. Nothing's changed. Yesterday's bad content is today's slop (plus a mind boggling amount of investment, corruption and environmental side effects).
I am assuming it is a bunch of manual work.
it already uses Workers Cache for the route-level ISR cache
I just don’t understand why undermine your own announcements by delegating comms to the machine. It’s disrespectful to the reader.
I found it funny to just read through.